JUST IN: Iran just bombed the only country willing to broker its peace.
Drones struck the Port of Salalah in Oman on 11th March, hitting fuel storage tanks at the MINA Petroleum Facility. Fires ignited. Then spread. As of tonight, the blaze has consumed most if not all oil tanks at the facility, burning into the darkness in a port that was not a military target, not an ally of the United States or Israel, but the neutral mediator that hosted the last diplomatic channel between Washington and Tehran as recently as February 2026.
Oman brokered the secret talks that led to the 2013 JCPOA framework. Oman hosted the February 2026 nuclear discussions that were the final diplomatic contact before 28 February. When every other Gulf state chose sides, Oman chose neutrality. When Iran needed a phone line to Washington, Oman was the phone. That phone is now on fire.
Iran’s response was extraordinary. President Pezeshkian called Oman’s Sultan and said the incident would be “investigated.” Iran’s military denied launching attacks on Oman, calling the suggestion a “false flag.” But the drone signature matches IRGC patterns. The fires are real. The fuel tanks are burning. And no other actor in the region has the capability, the reach, or the motive to strike Salalah with the drone systems that hit it.
This is the Mosaic Doctrine consuming its own creator’s diplomacy. The 31 autonomous IRGC provincial commands that operate without central authorisation do not consult Tehran’s Foreign Ministry before launching. A commander with coastal access to the Gulf of Oman can strike Salalah without knowing or caring that President Pezeshkian needs Sultan Haitham’s phone line to survive the war. The diplomatic wing of the Iranian state needs Oman alive. The military wing just set its oil tanks on fire. Both wings operate simultaneously without coordination because the doctrine was designed to make coordination unnecessary.
This is the structural impossibility nobody is modelling. Tomorrow, Larijani or Pezeshkian may call Muscat and beg forgiveness. They may ask Oman to reopen the channel to Washington. They may negotiate in good faith for a ceasefire. And while they are on the phone, an autonomous IRGC command in Hormozgan or Kerman may launch another drone at Salalah because the sealed orders from a dead Supreme Leader authorise continuous strikes on Gulf infrastructure and no living authority has the constitutional power to countermand them.
Peace requires trust. Trust requires that one side can guarantee what its own forces will do. Iran cannot guarantee what 31 independent commands will do because the man who could guarantee it is dead and his successor is a cardboard cutout. Oman cannot mediate between Washington and Tehran if Tehran’s military burns Omani infrastructure while Tehran’s president apologises for it. The mediator’s credibility dies the moment the mediator’s oil tanks ignite.
Salalah was the bypass. When Hormuz closed, shipping was supposed to reroute through Oman’s ports outside the Strait. When diplomacy was needed, Oman was supposed to carry the messages. When the war needed an off-ramp, Oman was supposed to build it. The IRGC just burned the bypass, silenced the messenger, and destroyed the off-ramp in a single night.
Iran’s economy runs on $5,000 per capita GDP, 60% inflation, and a currency that has lost 90% of its value under sanctions. It cannot afford to lose its only friend. It just did. And the doctrine that lost it was designed to be unstoppable.
https://t.co/eMrt5qYYst
You can now buy a single share of stock at the Nairobi Securities Exchange starting 1st August 2025.
Stocks will now trade in multiples of one — phasing out the Odd Lot Board.
Also, closing prices will only count if ≥100 shares trade in a session.
Men,
Remember,
When we gather together in the moonlit village gathering, it is not because of the moon.
Every man can see the moon in his compound.
We come together because it is good for men to do so.
~ Chinua Achebe
Here's what ChatGPT thinks:
Your thread is grounded in factual elements, though some points would benefit from clearer distinctions between correlation, causation, and emerging hypotheses. Let's unpack the factual basis and then explore why Kenyan writing may be more likely to be flagged as AI-generated:
---
1. Factual Basis of the Thread
Kenyan workers and AI training:
True. In early versions of ChatGPT, particularly GPT-3 and GPT-4, Kenyan workers (hired via platforms like Sama) were involved in labeling and moderating training data. Their work helped fine-tune model outputs and filter toxic content. Reports by Time (Jan 2023) confirmed this.
Wage rates of $2–$5/hr:
True. Public reports confirm these were the typical pay rates.
Worldcoin’s iris scanning in Kenya:
True. Worldcoin launched its biometric identity initiative in Kenya and was later suspended due to privacy concerns and government investigations.
Kenyan dominance in the academic ghostwriting industry:
Largely true. Kenya has become a global hub for academic ghostwriting, particularly in essay mills. Articles in The New York Times, Al Jazeera, and BBC have documented this industry over the past decade.
Impact of Kenyan writing style on AI models:
Partially speculative but plausible. While it’s not formally documented that Kenyan linguistic patterns specifically influenced model outputs, the idea that training data shapes outputs is core to how LLMs work. If enough data came from Kenyan-written or labeled content, traces of that style could influence outputs — though it's difficult to quantify without access to OpenAI's training sets.
---
2. Why Kenyan Work May Be Flagged as AI-Generated
Several intersecting factors could explain this:
a) Style Overlap with LLM Outputs
Kenyan academic writers are known for using grammatically correct, highly structured, formal English — which mimics the tone often found in AI-generated writing.
Many AI detectors use surface-level heuristics: coherence, lack of personal narrative, "over-formality", or certain sentence rhythms — all of which may match Kenyan academic writing, especially ghostwritten work.
b) Shared Training Sources
If LLMs were trained using publicly available academic databases, ghostwritten assignments from Kenyan writers (sold via essay mills) may have ended up in the data. This could create unintentional stylistic convergence.
c) Bias in Detection Tools
AI detection tools (like GPTZero or Turnitin's AI detectors) often rely on U.S.-centric benchmarks of what “human writing” should look like.
African English varieties (Kenyan English, Nigerian English, etc.) may appear anomalous to such tools — especially when formal but devoid of colloquial U.S. expressions.
d) Lack of Digital Fingerprinting
Kenyan writers working for clients abroad often write in polished, impersonal voices — lacking metadata or contextual signals (like location, device history, consistent voice) that could validate authorship.
---
In Summary
Your thread is largely factual with some reasonable extrapolation, particularly regarding influence on LLM outputs and AI detectors.
The higher likelihood of Kenyan writing being flagged as AI-generated comes down to a combination of stylistic similarity, possible overlap in training data, and biases in detection algorithms built on Western norms.
Applications are now open for our CBK Internship Program. Don’t miss the opportunity to gain workplace experience and grow your career. The deadline for submitting online applications is Friday, May 16, 2025. https://t.co/SbES2b2UTU
Good people, your girl here genuinely asking for help.Sitaki kukwama tena na hii laptop ni yetu 😂 Team retweets, let's do this! Retweet to 999 and I’ll finally get the laptop to kickstart my tech path, let's make it happen 🙏